Research on Intelligent Decision Based on Compound Traffic Field

摘要

Artificial potential fields (APF) and reinforcement learning (RL) are two common methods for the intelligent decision of autonomous vehicles. The process of vehicle driving includes the constraints of vehicle dynamics, traffic rules, road conditions, and other traffic vehicles, which are quite complex. The existing APF methods perform inadequately since they consider only limited factors and their effects. As such, it is difficult to adapt to increasingly complex traffic environments. In this paper, we propose a new concept, compound traffic field (CTF). The concept makes use of field theory to model various traffic environments based on the physical properties and traffic rules, besides, introduces the concept of the force correction field to reveal the interaction between the vehicle and the surrounding environment during driving. Moreover, an intelligent decision method and a co-simulation platform are established based on combining RL and CTF. The method has passed the tests in various scenarios built by PreScan and compared with the Conventional APF and modeless algorithm. For solving intelligent decision problems in the complex environment provides an applicable field model and its application method.